AMStraMGRAM: Adaptive Multi-cutoff Strategy Modification for ANaGRAM
Abstract
Recent works have shown that natural gradient methods can significantly outperform standard optimizers when training physics-informed neural networks (PINNs). In this paper, we analyze the training dynamics of PINNs optimized with ANaGRAM, a natural-gradient-inspired approach employing singular value decomposition with cutoff regularization. Building on this analysis, we propose a multi-cutoff adaptation strategy that further enhances ANaGRAM's performance. Experiments on benchmark PDEs validate the effectiveness of our method, which allows to reach machine precision on some experiments. To provide theoretical grounding, we develop a framework based on spectral theory that explains the necessity of regularization and extend previous shown connections with Green's functions theory.
Cite
@article{arxiv.2510.15998,
title = {AMStraMGRAM: Adaptive Multi-cutoff Strategy Modification for ANaGRAM},
author = {Nilo Schwencke and Cyriaque Rousselot and Alena Shilova and Cyril Furtlehner},
journal= {arXiv preprint arXiv:2510.15998},
year = {2025}
}